There seems to be some gray-area confusion about scanning piano rolls
and archiving said information.  Firstly, scanning the original paper
roll, whether optically or pneumatically, yields merely a picture of
the original roll.  The important step is taking this so-called picture
and converting it into the mathematical original (usually a stretched,
multiple-formula equation) created by the manufacturer that was
directly correlated to the mechanical advance of their perforating
equipment.

This is the only way to get an accurate copy of the production roll.
All information must be returned to original master/stencil format.
Even this information must be corrected and checked against master
stencils, if any exist.  Here at Keystone Music I have personally taken
optically scanned roll information, scanned and converted by Larry Doe,
one of the Keystone Team, corrected that information, and compared
hundreds of these conversion stencil scans against the original paper
masters, and I have found them to be exact.

I think the confusion about accuracy arises because different people
have different goals in making piano roll scans. I can only comment on
the process where the goal is to produce a paper music roll that is a
faithful copy of the original.  If the goal is to get a rendition of
the roll performance that plays audibly on a computer, then the
emulation program, not the scan, becomes all-important.  As Julian Dyer
recently posted, the sophistication of that program is what determines
the quality of the results.

Up to now the programs that I have heard are barely adequate in
capturing the information coded on reproducing rolls.  The mindset of
trying to exactly copy commercially issued piano rolls is an utter
waste of time for too many reasons to list here.  Re-creating the
company's corrected stencil master is the only way to produce a
satisfactory roll, having the coded information on that roll just like
the manufacturer intended it to be.

Richard Groman